GRADE concept article : Evaluating risk of bias in a body of evidence from studies of environmental and other exposures

UNSPECIFIED (2026) GRADE concept article : Evaluating risk of bias in a body of evidence from studies of environmental and other exposures. Environment international, 213: 110310. ISSN 0160-4120

Abstract

Evaluating the impact of bias in observational epidemiology studies is critical to systematic reviews in public health (as is the case with any study type). However, researchers have expressed concern that useful epidemiological data may not receive full consideration in some reviews due to overemphasizing potential for bias without critical assessment of its impact. This concern should be balanced against the need to appropriately assess the degree of bias in observational studies to accurately characterize their validity and reach appropriate conclusions. In this paper, we describe how examining bias across studies may alleviate or solidify concerns identified in individual studies. We describe how to assess the likelihood that specific sources of bias explain and/or meaningfully distort an observed association. We followed the methods for development of Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) workgroup guidance. Various sensitivity analyses are proposed to probe risk of bias in the body of evidence. Two illustrative case studies were conducted: the associations between trichloroethylene (TCE) exposure and kidney cancer occurrence and between two individual per-and polyfluoroalkyl substances (PFAS) and changes of alanine aminotransferase (ALT) levels. Sensitivity analyses can inform the potential for bias to influence the results of a body of evidence, including consideration of direction of bias, subsets/stratification, counterbalancing biases, use of positive/negative controls, and others. To incorporate these into a synthesis, consideration is needed throughout the planning and execution process: (1) Planning the systematic review; (2) Assessing risk of bias in individual studies; (3) Performing sensitivity analyses, including exploration of heterogeneity and bias impact; and (4) Determining final rating. These nuanced concepts and approaches for more accurately characterizing the magnitude and impact of bias in a body of evidence can be applied within GRADE and more broadly to systematic reviews outside that framework.